liveBook · Manning

Imitation learning

As explained in Applied Reinforcement Learning

Imitation learning trains an agent to reproduce behavior demonstrated by an expert or another policy.

Imitation learning trains a model by showing desired outputs and increasing the likelihood it will reproduce them. In supervised fine-tuning (SFT), curated query–response demonstrations teach domain-specific format, tone, and baseline behavior. The model imitates expert responses rather than exploring alternatives or optimizing a reward. section 10.5

Medical chatbot

For a patient asking whether tiredness and dizziness could mean low iron, a demonstration can model an empathetic answer, mention other possible causes, and recommend medical evaluation and testing. section 10.5

How Applied Reinforcement Learning covers it

How deep it goes
Explains expert demonstrations, output-likelihood imitation, and limits around choosing between acceptable responses.
Examples
Uses medical chatbot query-response pairs, including a question about fatigue, dizziness, and possible iron deficiency.
What you can do after
Evaluate imitation learning as a way to teach format, tone, and baseline behavior, not choose between acceptable answers.
How it connects
Contrasts imitation learning with exploration and assigns preference tradeoffs to reward modeling.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Applied Reinforcement Learning. The section links point to the parts of the book it draws on.